Data source
Health screening for the entire population is conducted by the National Health Insurance Service (NHIS) every year or every alternate year, depending on the age and occupation of people covered in the insurance and released to the public [17]. We conducted a longitudinal cohort study based on the general health screening database of the NHIS. The sampling research database was extracted by sampling high-demand information from the national health insurance data. We used the health examination cohort database to analyze the medical use and examination results. Approximately 510,000 people (aged 40-79 years) underwent routine checkups from 2002 to 2003. The data were extracted from January 01, 2002 to December 31, 2015.
This study was conducted with the approval of the institutional review board (IRB No. X-2004/606-905, Seoul National University Bundang Hospital).
Study population
We extracted the diagnosis code from the NHIS database, following the Korean Standard Classification of Diseases and Causes of Death-7 (KCD-7) code, a modified version of the 10th edition of the International Classification of Diseases and Related Health Problems. The treatment code was based on the treatment and surgery fee codes of Korea.
The inclusion criteria were women (i) aged over 50 years, (ii) who were administered BP for more than 1 year, and (iii) demonstrated compliance ≥ 0% during the year (or those with a medication possession ratio ≥ 80%). Both the oral and injectable routes were covered. Thus, the analysis eventually included BP, namely alendronate, ibandronate, residronate, pamidronate, and zoledronate.
However, the underlying condition or some contributing factor would have effects on the onset of ONJ, and the estimated ONJ frequency is quite higher in cases of cancer or other bone diseases that need to use BP. Therefore, the case of using BP above the dose for osteoporosis treatment, and the case of cancer were excluded.
In addition, we excluded cases where ONJ occurred within 1 year of the use of BP. This can be attributed to our difficulty in determining the result of BP use. Deaths before the index date were also excluded from the analysis.
Definition of ONJ, pulp and periapical disease, and caries
We combined the KCD-7 code and treatment code to diagnose ONJ. First, we categorized ONJ into two parts as follows: definite ONJ and possible ONJ. The condition was defined as definite ONJ if it included one of the following: Osteonecrosis due to drugs, multiple site (M87.1), Inflammatory conditions of jaws (K10.2) as a diagnosis, and one of the treatment were registered together, or both diagnoses were registered after the index date (Supplementary Table 1).
Possible ONJ is a broader concept that covered the diagnosis and treatment used in cases of suspected ONJ. It included Osteonecrosis due to drugs, multiple sites (M87.1, 87.3, 87.8, 87.9), Inflammatory conditions of jaws (K10.2, 10.3 10.8), Acute haematogenous osteomyelitis, multiple sites (M86), Disorders of teeth and supporting structures, unspecified (K08.9), as a diagnosis, and treatment code (Supplementary Table 2).
Pulp and periapical disease, and caries were defined as the diagnosis code in registration, following the index date (Supplementary Table 3).
Root canal treatment was defined as the root canal treatment code on record between pulp and periapical disease, or caries diagnosis and ONJ diagnosis (Supplementary Table 4).
Confounding variables
We included patient demographics on the baseline year, age, income level, and disability. The income quintile refers to an index in which the National Statistical Office divides Korea's total households into ten levels by 10%, according to the quarterly income level. The income level in the first quintile is the lowest. Moreover, the higher the quintile, the higher is the income. The analysis was divided into five stages. Disability was categorized as yes or no (Table 1, Supplementary Table 5). The prevalence of diseases was defined by the presence of hypertension (I10, I15), diabetes (E10, E118, E119, E13, E149), hyperlipidemia (E78), myocardial infarction (I21, I22), stroke (I60–63), or anemia (D46, D50-53, D55-64, D74) (Table 1, Supplementary Table 5).
Statistical analyses
Baseline characteristics were reported using numbers (%) for categorical variables and means ± standard deviation. We conducted the χ2 test, median with interquartile range for continuous variables, and t-test or analysis of variance. Moreover, we adjusted the cohort data of 14 years, baseline year, age, comorbidities, income disability, the type of BP administered, and cumulative proportion of days to analyze the hazard ratio (HR) of ONJ, according to the prevalence of pulp and periapical disease, caries, or pulp and periapical disease or caries. Cox proportional hazards models were used to evaluate the associations between the independent variables and the occurrence of ONJ events. In addition, we calculated the HRs and 95% confidence intervals (CIs) to evaluate the associations between the independent variables and the occurrence of ONJ events. Statistical significance was defined as two-tailed p-values <0.05. Statistical analysis was conducted using R programming (version 3.3.3; The R Foundation for Statistical Computing, Vienna, Austria).